This post is also available in:
עברית (Hebrew)
A research team from La Sapienza University in Rome has demonstrated a method for identifying individuals using nothing more than the way their bodies affect nearby Wi-Fi signals. The system, named WhoFi, can distinguish people with over 95% accuracy—without requiring them to carry any device.
Unlike conventional surveillance technologies such as facial recognition or location tracking via mobile phones, WhoFi works passively. It detects how a person’s presence alters Wi-Fi transmissions in the surrounding environment, using a standard technique known as Channel State Information (CSI). CSI captures detailed data on how signals propagate and change in real time, including both amplitude and phase shifts caused by interference from physical objects—such as the human body.
When analyzed by a deep neural network, these subtle signal distortions form a biometric “signature” that can be used to re-identify individuals as they move across different Wi-Fi zones. In this case, the researchers employed a transformer-based architecture, the same class of machine learning models that power many leading AI systems.
The research builds on earlier efforts to use wireless sensing for motion detection and person identification. Previous systems demonstrated the concept with limited accuracy. WhoFi, tested using the NTU-Fi dataset, pushes the technique much further, achieving a reported 95.5% match rate.
One key benefit of this approach is that it bypasses the need for cameras or active device tracking. It can function in low light, through walls, and without capturing visual images. However, the technology also introduces new ethical concerns about covert surveillance and the potential for passive tracking without consent.
As Wi-Fi sensing capabilities continue to advance, the line between convenience and surveillance becomes increasingly blurred. While systems like WhoFi may offer promising applications in security, smart environments, and healthcare, they also raise important questions about consent, oversight, and how biometric data is collected and used. As adoption grows, policymakers and technologists alike will need to consider how to balance innovation with individual privacy in an era where simply walking through a room may be enough to leave a digital trace.
The findings of the study were published as a preprint on arXiv.
